Marine Geology 396 (2018) 160–170

Contents lists available at ScienceDirect

Marine Geology

journal homepage: www.elsevier.com/locate/margo

Scenario of the 1996 volcanic in Karymskoye Lake, Kamchatka, T inferred from X-ray tomography of heavy minerals in deposits

⁎ Simon Falvarda, Raphaël Parisa, , Marina Belousovab, Alexander Belousovb, Thomas Giachettic, Stéphanie Cuvend a Université Clermont Auvergne, CNRS, IRD, OPGC, Laboratoire Magmas et Volcans, F-63000 Clermont-Ferrand, France b Institute of Volcanology and Seismology, Petropavlosk-Kamchatsky, Russia c Department of Earth Sciences, University of Oregon, Eugene, USA. d IODP-France, 14 Av. Edouard Belin, 31400 Toulouse, France.

ARTICLE INFO ABSTRACT

Keywords: The concentration and distribution of heavy minerals in tsunami deposits is not random and mostly source- Tsunami dependent. Heavy minerals may thus be good indicators of sediment provenance and tsunami flow dynamics. Phreato-magmatic eruption The tsunamis generated by the 1996 phreato-magmatic eruption in Karymskoye Lake represent a relevant case- Heavy minerals study because the provenance of the abundant heavy minerals found in the tsunami deposits is well constrained X-ray tomography (the on-going basaltic eruption itself). X-ray computed tomography (X-CT) of cores of tsunami sediments is used Karymskoye Lake to identify heavy minerals and characterise their source and spatial distribution in the tsunami deposit, and to propose a scenario of the coupled eruption and tsunamis. An original combination of methods including X-CT, SEM and XRF core scanner allows distinguishing subunits corresponding to pulses of sediments deposition and associated inputs of heavy minerals, together with erosive contacts, laminations, and rip-up clasts of the substratum. The structure of the tsunami deposits suggests that a major tsunami consisting of two main waves inundated the coastal terrace up to 100 m inland on the eastern shore of the lake; a scenario that is consistent with waves generated by experimental explosions. This largest tsunami might have occurred when underwater explosions were at a critical water depth of 40 m (corresponding to a two-third submerged explosion in the 60 m deep lake). However, more investigations are needed to better understand the critical conditions leading to a tsunami during underwater eruptions.

1. Introduction concentrations occurred towards the upper part of the 2004 tsunami deposits in Thailand, with mica flakes found more commonly in the Heavy minerals in clastic sediments and sedimentary rocks are finest-grained sediment samples. However, the 2011 Tohoku-oki tsu- commonly used to study diagenesis, weathering processes, sediment nami deposits in Sendai, which are particularly rich in heavy minerals, provenance, depositional processes, and to reconstruct paleoenviron- show no different assemblage compared to other pre-tsunami sediments ments (e.g. Dill, 1998; Mange and Wright, 2007). Heavy minerals are and no vertical trends with a 1 cm-interval sampling (Jagodziński et al., often reported in tsunami deposits (e.g. Switzer et al., 2005, 2012; 2012). Predominance of local sediments (beach, dune, soil) upon Bahlburg and Weiss, 2006; Szczuciński et al., 2006; Babu et al., 2007; offshore sediments is confirmed by the rarity of marine bioclasts in Morton et al., 2007, 2008; Narayana et al., 2007; Higman et al., 2008; the 2011 tsunami deposits at Sendai (Pilarczyk et al., 2012; Szczuciński Srinivasalu et al., 2008; Jagodziński et al., 2009, 2012; Moore et al., et al., 2012). 2011; Costa et al., 2015). A higher concentration of heavy minerals was The concentration and distribution of heavy minerals in tsunami estimated in tsunami deposits (Jagodziński et al., 2009) and post- deposits is mostly source-dependent, but they are promising indicators tsunami deposits (Babu et al., 2007) sediments, compared to pre- of sediment provenance and flow dynamics. Cuven et al. (2013) tsunami deposits. Heavy minerals described in recent tsunami deposits concluded that “further investigations are needed to improve under- are likely concentrated in pockets or laminae (e.g. Morton et al., 2007; standing of the spatial distribution of different populations of grains, in Higman et al., 2008; Srinivasalu et al., 2008; Jagodziński et al., 2012; terms of density and shape, and their relationship to sediment sorting in Switzer et al., 2012). Jagodziński et al. (2009) noted that higher mica tsunami deposits”. In this new contribution, X-ray tomography is used

⁎ Corresponding author. E-mail address: [email protected] (R. Paris). http://dx.doi.org/10.1016/j.margeo.2017.04.011 Received 28 October 2016; Received in revised form 21 April 2017; Accepted 23 April 2017 Available online 28 April 2017 0025-3227/ © 2017 Elsevier B.V. All rights reserved. S. Falvard et al. Marine Geology 396 (2018) 160–170 to (1) identify heavy minerals and characterise their source and spatial Akademii Nauk caldera to Karymsky stratovolcano, one of the most distribution in a tsunami deposit, and (2) to propose a scenario of the active volcanoes in Kamchatka. 1996 volcanic tsunamis in Karymskoye Lake (Kamchatka, Eastern After a strong on January 1st, 1996, there was an Siberia, Russia). This tsunami is a good candidate because the heavy intense seismic swarm and two eruptions started simultaneously from mineral population is mostly provided by the sublacustrine volcanic Karymskoye Lake and from the central vent of Karymsky volcano on eruption that generated the tsunami. The approach might give some January 2nd (Muravyev et al., 1998). Phreatomagmatic activity formed insights on the coupling between the eruption and the tsunami(s) and a new tuff-ring attached to the northern slope of the lake. The ice understanding of wave generation during phreatomagmatic eruptions. covering the lake might have prevented the generation of tsunami Volcanic tsunamis are generated by a variety of mechanisms, including waves during the first stages of the eruption. Pyroclastic base surges volcano-tectonic , slope instabilities on volcano flanks, damaged bushes as far as 1.3 km from the vent in all directions, at pyroclastic flows, underwater explosions, shock waves, and caldera elevations up to 150 m above the level of the lake (Belousov and collapse (e.g. Paris, 2015). Among the historical record, volcanic Belousova, 2001). Ash fall deposits extend mostly to the southeast, with tsunamis represent a low-frequency (about 5% of all recorded a thickness decreasing from 5 cm at 1.5 km from the vent to 1 cm tsunamis) but four events (Ritter Island 1888, Krakatau 1883, Unzen at > 3 km from the vent on the shores of the lake. Large ballistic blocks 1792, and Oshima-Oshima 1741) are ranked in the twenty deadliest of juvenile basalt, hydrothermally altered rocks and rhyodacitic pumice volcanic disasters (Auker et al., 2013). were ejected (Grib, 1998; Belousov and Belousova, 2001). Observation of empty bomb sags in September 1996 was interpreted as an evidence 2. The 1996 eruption and tsunamis in Karymskoye Lake of ice blocks ejected during initial explosions (Belousov and Belousova, 2001). Karymskoye Lake fills the late Pleistocene Akademii Nauk caldera, Tsunamis generated by underwater explosions were observed dur- fl fl the later belonging to the Eastern Volcanic Belt of Kamchatka ing helicopter ight and periodically triggered debris ows in the Peninsula. The lake is 4 km across, with steep flanks and a flat bottom Karymskaya River (Muravyev et al., 1998). Belousov and Belousova 3 ff at 60–65 m deep (Fig. 1). This ~0.5 km volume of fresh water is fed by (2001) described the e ects of the tsunamis on the shores of the lake, several streams and hot springs, and drained by the Karymskaya River and Belousov et al. (2000) estimated wave runups. The values of runup to the North (Belousov and Belousova, 2001). The post-caldera volcanic decrease with the distance from the source, i.e. the eruptive vent, with a history is poorly documented and apparently limited to scoria cones maximum runup of 19 m on the north-western shore (< 1 km from the and tuff rings on the northern shores of the lake (Belousov et al., 1997). vent) and a minimum runup of 1.8 m on the south-eastern shore (3 km ff 3 Eruptive vents are aligned along a North-South linking the from the vent). Boulders of volcanic tu up to 1.3 m were transported

Fig. 1. Location map of Karymskoye Lake, Kamchatka Peninsula, Russia. Reconstructed pre-eruption bathymetric contours (modified from Torsvik et al., 2010) are in meters below the surface of the lake (black bold line at 624 m a.s.l.). Red circle indicates the location of the tuff-ring formed during the 1996 eruption. Deposits of the 1996 tsunami were studied all around the lake (black squares), and the sections sampled for X-ray tomography are indicated by yellow squares (K02 and K08). Geographical coordinates in decimal degrees. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

161 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 2. Effects of the 1996 eruption and tsunamis on the northern shore of Lake Karymskoye. Note the formation of a new tuff-ring with an inner diameter of 500 m, extensive destruction of the vegetation (bush) and erosion of the shoreline, formation of new beaches, and slope erosion near the eruptive vent. Modified from Paris (2015). Photograph showing the volcanic ash plume generated by one of the phreatomagmatic explosions (courtesy of Ya. D. Muravyev). up to 60 m inland. Soil erosion by successive tsunamis led to the underwater explosion of 5 × 1014 J generating a water cavity with a formation of talus up to 2–3 m high all around the lake, except on the diameter of 350 m and an initial surface displacement of 55 m can southern rocky shore. Sediment reworked by tsunami waves left new explain the observed runups of the largest tsunami. beaches up to 50 m wide and sandy patches on terraces behind the talus (Fig. 2). These tsunami deposits are often > 1 m thick on the beach, 3. Methods but < 30 cm on the terraces and gentle slopes, and display a complex inner structure. The mean grain size of the tsunami deposits (from fine Survey on the shores of Karymskoye Lake (Kamchatka, Russia) was to coarse sands) is similar to distal base surge and fallout deposits carried out in August 2009. Tsunami deposits were retrieved in plastic produced by the volcanic eruption, but tsunami deposits have a wider boxes (blocks) and carbon tubes (cores), thus preserving their structure range of sorting values (Belousov and Belousova, 2001). Deposits (Fig. 3). Bulk samples were collected along trenches for grain size ff fi forming the emerged part of the tu -ring are signi cantly less sorted analysis (laser diffraction on H2O2 treated sediment) and composition and coarser-grained (coarse sand to pebble) than the other deposits. (binocular). Thin-sections were made for each box sample using The limit of inundation can be traced by drift wood and rounded dehydration by water-acetone-epoxy-exchange, and analysed with a pumice lapilli (coming from old rhyolitic pyroclastic deposits of the SEM (JEOL JSM-5910, Laboratoire Magmas & Volcans, Clermont-Fer- caldera). rand, France). Chemical analyses were performed on dehydrated blocks Belousov and Belousova (2001) argued that a large tsunami (larger of sediments using an Avaatech X-ray Fluorescence (XRF) core scanner than the ones observed from helicopter) occurred at the end of the (EPOC laboratory, Bordeaux). High-resolution chemical relative ele- eruption, mostly because the tsunami deposits and areas eroded by ment variations were captured at the surface of the sediment every tsunami waves are not covered by pyroclastic fall deposits. The wave 500 μm with a ionization energy of 10 kv. runups measured all around the lake would then correspond to this X-ray Computed Tomography (X-CT) of the tsunami sediments in tsunami. Numerical simulations of underwater explosions and related carbon tubes was realised using (1) an Explore CT-120 GE Healthcare tsunamis in the lake (Ulvrová et al., 2014) demonstrated that an (INSERM Clermont-Ferrand) at a resolution of 50 μm per voxel, and (2)

162 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 3. Example of sampling at site K02-9 (eastern shore of the lake). Bulk sampling (white dots) of the 1996 tsunami deposit (K02-9A to K02-9D) and pre-tsunami sediments (K02-9E). The tsunami deposit is also sampled using carbon tubes (cores for X-ray tomography) and plastic boxes (for thin sections). Note the 20 cm-large rip-up clast of white ash at the base of the tsunami deposit (on the left). the ESRF beamline ID19 (Grenoble, France) at a resolution of 15 μm per coarse clasts prevented sampling K02-8, K02-5 and K02-6 in carbon voxel (one sample only: K02-9). X-CT is an analytical technique based tubes. Dehydrated blocks and thin sections were prepared for samples on the absorption of X-rays by the sample (Hounsfield, 1962). This K02-9 (upper 10 cm only) and K02-8. Results obtained on the eastern means that particles of different densities will appear in different shades shore of the lake are compared with a proximal section on the western of grey on the images. Due to their high density, heavy minerals appear shore (K08), at 1 km from the source of the tsunamis, i.e. the eruptive in very bright grey to white tones on the X-CT images, which allowed to vent (Fig. 1). easily separate them from the other particles by using a simple threshold value. Particles with grey values corresponding to the heavy 4. Results minerals were turned white and particles and matrix of lower density were turned black. The choice of the threshold value was validated by 4.1. Characteristics of tsunami deposits on the eastern shore determination of the minerals on thin sections (mostly olivines, clinopyroxenes and Fe-Ti oxides). The position of each heavy mineral Preserved thickness of the 1996 tsunami deposits on the eastern in the core was then automatically measured and saved along 2D shore progressively decreases landward (Fig. 5), from 43 cm on the vertical slices of the cores. Heavy minerals abundances along the cores beach (K02-9) to 3 cm at 30 m from the shoreline (K02-6). This were calculated using a 1 mm subsampling step with a Matlab code (see landward thinning is associated with landward fining of the mean Falvard and Paris, 2016, for more details). The higher resolution of the grain size, but the tsunami deposits are internally structured in subunits scans performed at ESRF ID19 (15 μm per voxel) allowed measuring the with different characteristics. The contact with the underlying soil and intermediate diameter (B-axis) of each particle along a vertical slice of riverbed sediments is always erosive. Rip-up clasts of soil and white the sample, plotting a map of the grain size distribution and measuring compacted ash are frequent (Fig. 3). Thickest section K02-9 displays 4 GSD (Grain Size Distribution) parameters (using Matlab codes proposed subunits (A-D). Basal unit (K02-9D) is a massive 23 cm thick dark by Falvard and Paris, 2016). medium sand with > 50% of fresh (juvenile) basaltic pyroclasts and For this study, we focus on 9 trenches along a 25 m long transect on crystals (plagioclases, olivines and clinopyroxenes) from the 1996 the eastern shore of the lake (K02 on Fig. 1). Belousov et al. (2000) eruption. There is no other possible source of such an assemblage of measured runups between 1.8 and 2.4 m on the eastern shore of the minerals in the local stratigraphy. Most frequent texture of juvenile lake. Low topography of this area favoured penetration of tsunami basaltic clasts found in the K02 sections is blocky and poorly vesicu- waves inland, up to 100 m, even if the wave heights were small lated. Texture of the basaltic clasts sampled in the tuff ring ranges from compared to the other shores. Transect K02 is oriented N135°, i.e. poorly to highly vesiculated, with occasional veins of rhyolite (Grib, perpendicular to the shoreline and aligned with the eruptive vent 1998; Belousov and Belousova, 2001). The presence of rip-up clasts of (Fig. 4). The first ten meters of the transect (trenches K02-8 and K02-9) soil at the base of the subunit and crude laminations observed on X-CT correspond to the new beach formed during the 1996 eruption and imagery excludes primary fallout deposition. Subunit KO2-9D has high tsunamis. The shores of the lake were vegetated by alder bush before concentrations of heavy minerals, with two pulses at > 1500 heavy the tsunami. Trench K02-7 is on a gentle slope break between the beach minerals per mm (Fig. 5). and a terrace. The other trenches (K02-1 to K02-6) are located on this Second subunit 9C scours subunit 9D and a 20 cm-large rip-up clast terrace that is 1–2 m above the level of the lake (624 m a.s.l.). Roots or of white ash marks the contact (Fig. 3). Subunits 9C, 9B, and 9A are all

163 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 4. Longitudinal transect (KO2) in the 1996 tsunami deposit on the eastern shore of Karymskoye Lake. The tsunamis were generated by underwater explosions from the tuff-ring (dotted circle). X-ray tomography was performed on 6 core samples. laminated: horizontal to weakly oblique laminae of coarse silt to fine Despite its small size and smooth morphology, the talus between the sand in 9C, crudely laminated fine to coarse sand in 9B, and cross- beach and the terrace (Fig. 4)affected the depositional processes, as laminae of very fine to medium sand in 9A (Figs. 6 and 7). Similar evidenced by the structure and texture of tsunami deposits attached on laminations are observed on section K02-8. Compared to the dark sand the slope. There are no longer distinct subunits visible on section K02-7, of basal subunit 9D, the other subunits have less 1996 juveniles which is obliquely laminated and inversely graded from very fine sand (< 40%) and less heavy minerals (< 700 per mm), but more pumice at the base to medium sand at the top. Inverse grading is coupled with (> 20%) and rare diatoms in subunit 9A. The distribution of pumice an upward increase of the concentration of heavy minerals (Fig. 5). clasts in the tsunami deposits is not homogenous and their size ranges Lamination is dipping downward, the angle being similar to the from coarse ash to fine lapilli. Sublayers or lines enriched in pumices topography of the talus. (40–90%) coinciding with peaks of Si/Ca were identified on thin On the terrace but still close to the talus, section K02-1 is poorly sections and X-CT (Fig. 8). The concentration of heavy minerals (X- organised and no internal structure or vertical grading could be CT) is not grain-size dependent (Fig. 6) and correlates with peaks of Fe/ detected on the field. However, X-CT allows distinguishing two Al and Ca/Al, which represent respectively higher concentrations of successive pulses of medium-to-coarse sand particularly rich in heavy oxides (mostly titanomagnetites) and olivines, and clinopyroxenes. minerals, separated by a horizontal laminae of very fine sands with less

0 0 0 0 0 0 A A A A −50 −50 −50 −50 −50 −50 A B B B B 05001000 1500 2000 −100 −100 −100 −100 number of HMs per mm 0 500 1000 1500 2000 2500 3000 0 500 1000 B number of HMs per mm C number of HMs per mm

−150 −150 −150 0 500 0 500 1000 number of HMs per mm number of HMs per mm −200

D

−250

−300 0 500 1000 1500 2000 depth (mm) number of HMs per mm

Fig. 5. Concentration of heavy minerals (HMs) inferred from X-CT on cores of tsunami deposits. See Fig. 4 for location of the sections.

164 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 6. Concentration of heavy minerals (X-CT, units per mm) compared to grain size inferred from image analysis of X-CT (MSi: medium silt; CSi: coarse silt; VFSa: very fine sand; FSa: fine sand; MSa: medium sand), and major element chemistry (XRF) of the upper part of section K02-9 (subunits A, B and C). The maximum values in Fe/Al and Ca/Al profiles represent respectively higher concentrations of oxides (titanomagnetite) + olivines, and clinopyroxenes. heavy minerals (at −50 mm on Fig. 5). Similar structure is found on clinopyroxene. Similar stratigraphy was found on other trenches the following sections (K02-2 to K02-6), with lower concentrations of located along the eastern shore of the lake (Fig. 1). In this area, terraces heavy minerals at K02-2 (< 500 per mm) and higher concentrations at higher than 2 m systematically prevented the deposition of sediments the base of sections K02-3 and K02-4 (> 1500 per mm). High by tsunami waves further inland. Note that the upper limit of tsunami concentrations of heavy minerals are associated with high proportions deposits coincides with the upper limit of driftwood. Thus, the absence (> 50% at the binocular) of juvenile material from the 1996 eruption of tsunami deposits does not mean that they were not preserved at such as basaltic pyroclasts and phenocrysts of plagioclase, olivine, and elevations higher than 2 m above the level of the lake.

165 S. Falvard et al. Marine Geology 396 (2018) 160–170

N 135 1 mm Although being located at 1 km from the eruptive vent and near the limit of pyroclastic surges, 1996 tsunami deposits of the western shore have always < 30% of juvenile clasts from the 1996 eruption.

5. Discussion

5.1. Distribution of heavy minerals in the tsunami deposits

X-CT proved to be a useful technique to detect heavy minerals in the tsunami deposits, with measured abundances ranging from several dozens to 3000 units per millimetre in the samples (Fig. 5), and despite a limited resolution (50 μm for INSERM scans). Note that 2D analysis could have been done using a mosaic of SEM (Scanning Electron Microscope) images, but X-CT allows working on selected horizontal or vertical slices representing the entire core or specific areas of interest. Comparing grain size data and heavy mineral abundances in the upper part of K02-9 (Fig. 6) shows that the distribution of heavy minerals along the deposit is not random and not grain size dependent. Subunits enriched in heavy minerals are clearly evidenced by the X-CT data. a These remarks are also valid for the pumice clasts (Fig. 8), whose behaviour in water waves is probably different from heavy minerals (except micas in a certain extent due to their flaky shapes, but none were found in these deposits). However, heavy minerals in the 1996 tsunami deposits are never distributed along thin laminae, as often described in other tsunami deposits (e.g. Morton et al., 2007; Higman et al., 2008; Srinivasalu et al., 2008; Jagodziński et al., 2012; Switzer et al., 2012). The formation of heavy minerals laminae requires certain sediment concentration and flow regime conditions (e.g. Cheel, 1990) that were apparently not fulfilled during the 1996 Karymskoye Lake tsunami. Distribution of heavy minerals together with the structure and texture of the tsunami deposits of the lower sections (e.g. K02-9 and K08), confirm that multiple tsunamis were produced during the volcanic eruption. This succession of tsunamis is evidenced by multi- bedded deposits (Fig. 9) whose distribution is limited to the beach (< 1.5 m above the level of the lake). Orientation of the laminations in the upper part of section K02-9 (Figs. 6 and 7) indicates transitions from b upper plane beds (horizontal lamination) to low-angle beds dipping

Fig. 7. Cropped views of sample K02-9 scanned at ESRF ID19. The resolution is 15 μm/ landward (dune-like bedforms). Capping subunits are often cross- voxel. Heavy minerals are in bright white and thus easy to identify throughout the bedded (Figs. 6 and 10), with either backset orientation if we consider deposit. a — Subhorizontal lamination in the intermediate part of the deposit (subunit that they were deposited by an uprush flow of the upper regime, or K02-9C). Almost invisible to the naked eye, laminations appear clearly on the X-ray foreset orientation if they correspond to a final backwash or drainout of imagery due to the low density of pumices intercalated between sand grains. Note the the lake. In any case, there is no clear relation between the abundance fi — fi ne-scale vertical grading in each laminae. b Landward plunging strati cation in the of heavy minerals, the vertical grading and the laminations (Figs. 6 and upper part of the deposit (subunit A), mostly composed of coarse pumices (notice the vesiculation inside the pumice clasts) and large volumes of porosity between layers of 10). finer sands and silts. However, X-CT reveals two peaks of heavy minerals in subunit K02- 9D (Fig. 5), as well as in all sections located on the terrace of the eastern shore (Fig. 5: upper sections K02-1 to K02-4). This transect-scale 4.2. Characteristics of tsunami deposits on the western shore correlation suggests that two main waves inundated the terrace up to 100 m inland on the eastern shore of the lake. This phase of inundation High terraces or steep slopes of the western side (K08) also limited corresponds to the maximum runup values measured by Belousov et al. inundation and deposition to the beach, where the 1996 tsunami (2000). The tsunamis observed during the helicopter flight (Muravyev deposits are particularly thick (commonly > 50 cm). Belousov et al. et al., 1998) did not inundate the terrace of the eastern shore. A larger (2000) reported wave runups higher than 5 m in this area. Due to a tsunami, either consisting of two main waves or an uprush-backwash different topographic setting and closer distance to the source, the succession, is thus recorded in the distribution of heavy minerals along structure of the 1996 tsunami sequence is different. Section K08 the upper sections. displays a complex succession of laminated sands and silts characterised by abrupt variations of grain size (Fig. 9). The base of the tsunami 5.2. A scenario of the volcanic eruption and tsunamis deposit is unclear and > 7 subunits can be distinguished. Low-resolu- tion X-CT on a dehydrated block allows identifying cross-laminae in the It is worth reminding that the explosions were not all tsunamigenic, upper part of the sequence (subunit K08-B on Fig. 8). This cross- as they simply produced a single, near-vertical, column-like jet of gas bedding is not associated to any particular distribution of heavy and pyroclasts that collapsed back into the lake, thus generating a minerals (Fig. 10). The deposits sampled along the western shore of subtle base surge. Initial stage of volcanic activity started at a water the lake (e.g. K08) are particularly rich in diatoms. The distribution of depth of 45–50 m when the surface of the lake was still covered by a diatoms is not random, and fine-grained sub-layers or lines enriched in 1 m thick icecap (Fig. 12A). Thus, the first explosions could not diatoms can be seen intercalated between coarser subunits (Fig. 11). generate tsunamis. During a transitional stage (Fig. 12B), the eruption

166 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 8. Distribution of pumices estimated from image analysis of thin sections at the SEM and Si/Ca profile inferred from XRF scan of dehydrated block (sample K02-8).

Fig. 9. Overview of section K08 (western shore of the lake). The 1996 tsunami deposit represents the entire thickness trenched and is organised in distinct subunits, which are all laminated (except subunit K08-E and lenses of coarse sand in K08-C). Note the rip-up clasts of soil and ash on the right side and the backset cross-laminae in subunit K08-B (X-CT).

167 S. Falvard et al. Marine Geology 396 (2018) 160–170

shell of water up to 450 m high (Muravyev et al., 1998; Belousov and Belousova, 2001). This mass then became unstable and was pierced by cypressoid jets of pyroclasts with gas and water. Simultaneously, a bore-like elevation of the water surface started to propagate radially at − about 20–40 m·s 1 thus forming a tsunami in the lake (Muravyev et al., 1998; Belousov and Belousova, 2001)(Fig. 12D and E). There is no ice anymore on photographs taken by Muravyev et al. (1998) three hours after the beginning of the eruption. This description represents the mature stage of the eruption (Fig. 12E), observed during the helicopter flight. There is no observation available for the initial and final stages. Multiple tsunamis and subsequent reflected waves were generated during the mature stage of the eruption, as evidenced by multi-bedded tsunami deposits of the lower sections (Figs. 3 and 9). The distribution of heavy minerals in the upper sections (Figs. 4 and 5: terrace along the eastern shore) records two successive waves corresponding to the maximum runup. It is interesting to note that tsunamis generated by underwater explosions consist of two main waves followed by smaller undulations, as demonstrated by experi- ments and numerical simulations (e.g. Le Méhauté and Wang, 1996; Kedrinskii, 2005; Torsvik et al., 2010; Ulvrová et al., 2014). The water is initially pushed upward and radially, forming a crater and a cylindrical bore expanding radially to form the leading wave, followed by a wave trough. The water crater then collapses as in a dam break model, thus generating a peak of water in the centre and a second cylindrical bore that progressively turns to a non-dissipative wave (Le Méhauté and Wang, 1996). However, a reflected wave or an uprush- backwash succession could also explain the structure and two-peak distribution of heavy minerals. Heavy mineral patches or shadows displaying a downstream decreasing abundance of heavy minerals in the second peak would have supported the backwash hypotheses (Cheel, 1984). However, there are no clear longitudinal trends of increasing or decreasing abundance of heavy minerals along the K02 transect (Fig. 5). It is actually difficult to state on the origin and precise timing of the largest tsunami (Fig. 12D). The poor quality of the videos taken from the helicopter does not allow identifying traces of erosion on the shores of the lake. Belousov and Belousova (2001) proposed that the largest tsunami occurred at the end of the eruption because its deposits are never covered by ash fallout deposits. Fragments of a hut originally located on the northern shore of the lake (and destroyed by pyroclastic surges and tsunamis) were found in drift wood left by the largest tsunami on the terrace of south-eastern shore. The transport of these debris from one side of the lake to another requires a significant time lapse between the destruction of the hut and the largest tsunami. However, it is difficult to imagine how explosions might have generated large waves when the tuff ring was already well developed and emerging (Fig. 12E). Experimental studies of underwater explosions (e.g. Craig, 1974) demonstrated that maximum wave heights occur for a two-third submerged explosion (i.e. at 40 m water depth in the case of Fig. 10. Heavy minerals (in red) and cross-bedded laminae of the upper part of the K08 section (see Fig. 9), western shore of the lake. Karymskoye Lake). Wave heights at this upper critical depth are twice that observed for explosions at the lower critical depth (fully sub- merged, i.e. at 60 m water depth).The eruptive phase corresponding to progressively broke the ice surface and fallout pyroclasts started to two-third submerged explosions might have generated the largest accumulate on ice rafts. Water expulsed by the explosions might have tsunami recorded in the stratigraphic sections and distribution of heavy surged and redistributed pyroclasts on the surface of the icecap before minerals (Fig. 12D). its complete melting. The first tsunamis then pushed ice rafts covered with pyroclasts on the shores of the lake (Fig. 12C). For areas which 6. Conclusion were not affected by pyroclastic surges and covered by < 3 cm of ash (e.g. eastern shore), this hypothesis could explain the unexpected high This study confirms that (1) X-CT combined with other methods proportion of juvenile basaltic clasts (and heavy minerals) in the thick such as XRF and SEM is a powerful technique for characterising the basal subunits of tsunami deposits (Figs. 3 and 5). Distal pyroclastic structure and composition of tsunami deposits (May et al., 2015; surge and fallout deposits settling in the lake between each explosion Falvard and Paris, 2016), and (2) provides another evidence of volcanic might have represented an additional source of heavy minerals for the eruption partially reconstructed using its tsunami record (Paris et al., tsunami deposits. 2014). X-CT helped evidencing the structure and characterising the Strongest observed explosions started with a rapidly rising (~3 s), distribution of heavy minerals in the 1996 tsunami deposits. In this case smooth-surfaced bulbous mass of gas and pyroclasts expanding within a study, the source of the heavy minerals is controlled by volcanic pulses

168 S. Falvard et al. Marine Geology 396 (2018) 160–170

Fig. 11. Diatoms in tsunami sample K08 viewed at the SEM and in thin sections. The distribution of diatoms is not random, and fine-grained sub-layers or lines enriched in diatoms can be seen at the base of coarser subunits.

W E eruption under ice tephra on ice ° ° ° ° ° ° ° ° lake a b

pyroclastic surge large tsunami tsunami ice rafts c d

tsunami tuff ring e f

Fig. 12. A scenario of the 1996 volcanic tsunamis in Karymskoye Lake. The lake is 4 km large. Vertical scale exaggerated. a — Initial stage of the phreatomagmatic eruption: a thick ice cap covers the whole surface of the lake, preventing any tsunami to occur; b — Emergent eruption: ice cap is pierced by the explosions, and basaltic pyroclasts (rich in heavy minerals) are scattered on the ice by fallout; c — Transitional stage: melting of the ice allows tsunami waves to propagate in the lake. Ice rafts are pushed towards the shores; d — Mature stage: largest tsunami occurs as an optimal ratio of explosion depth vs. water depth is reached (i.e. two-third submerged explosions); e — Mature stage (as observed during helicopter flight): the growing volcanic edifice gets closer to the surface of the lake and thus hampers the energy transfer from the explosion to the lake, preventing large waves to form; f — Final stage: fully formed tuff ring at the end of the eruption.

169 S. Falvard et al. Marine Geology 396 (2018) 160–170

(phreatomagmatic explosions) and it was possible to distinguish a large I., 2013. High-resolution analysis of a tsunami deposit: case-study from the 1755 Lisbon tsunami in southwestern Spain. Mar. Geol. 337, 98–111. tsunami composed of two main waves corresponding to peaks of heavy Dill, H.G., 1998. A review of heavy minerals in clastic sediments with case studies from mineral abundances. the alluvial-fan through the nearshore-marine environments. Earth Sci. Rev. 45, – However, it is still difficult to understand the critical conditions 103 132. Falvard, S., Paris, R., 2016. X-ray tomography of tsunami deposits: towards a new leading to a tsunami during underwater eruptions and to define hazard depositional model of tsunami deposits. Sedimentology 64 (2), 453–477. scenarios for future eruptions of the same style. Magma-water interac- Freundt, A., Strauch, W., Kutterolf, S., Schmincke, H.-U., 2007. Volcanogenic tsunamis in lakes: examples from Nicaragua and general implications. Pure Appl. Geophys. 164, tions during phreatomagmatic eruptions are complex and their physics 527–545. is controlled by water depth, geometry of the vent and magma-water Grib, E.N., 1998. Petrology of ejecta discharged by the January 2–3 eruption in the – interface, transfer of thermal energy, processes of intermingling and Akademii Nauk caldera. Vulkanol. Seismol. 5, 71 97. Higman, B., Bourgeois, J., Shiki, T., 2008. Deposits of the 1992 Nicaragua tsunami. In: mixing between magma and water, metastability of superheated water, Tsuji, Y., Yamazaki, T., Minoura, K. (Eds.), Tsunamiites – Features and Implications. and quantity of gas in the ascending magma (Kokelaar, 1986; Wohletz, Elsevier, pp. 81–103. Hounsfield, G.N., 1962. Computerized transverse axial scanning (tomography): part 1. 1986; Morrissey et al., 2010). Explosive eruptions accompanying the Description of system. Br. J. Radiol. 46 (552), 1016–1022. formation of maars and tuff rings, as occurred at Taal in 1965 and Jagodziński, R., Sternal, B., Szczuciński, W., Lorenc, S., 2009. Heavy minerals in 2004 Karymskoye in 1996, are characterised by phases of optimal ratios of tsunami deposits on Kho Khao Island, Thailand. Pol. J. Environ. Stud. 18, 103–110. Jagodziński, R., Sternal, B., Szczuciński, W., Chagué-Goff, C., Sugawara, D., 2012. Heavy (a) water vs. magma and (b) explosion (fragmentation) depth vs. water minerals in the 2011 Tohoku-oki tsunami deposits — insights into sediment sources depth. These explosions are able to generate tsunamis, whereas most of and hydrodynamics. Sediment. Geol. 282, 57–64. the Surtseyan-type explosions are not (Kokelaar, 1986; Paris, 2015). Heavy minerals in use. In: Mange, M.A., Wright, D.T. (Eds.), Developments in Sedimentology. 58 Elsevier (1283 pp.). Underwater eruptions represent < 10% of all recorded eruptions (if Kedrinskii, V.K., 2005. Hydrodynamics of Explosion. Springer, Berlin, Heidelberg. we exclude eruptions at mid-ocean ridges) but they caused 20% of all Kokelaar, P., 1986. Magma-water interactions in subaqueous and emergent basaltic volcanism. Bull. Volcanol. 48, 275–289. fatalities related to volcanic activity in historical time, including Le Méhauté, B.L., Wang, S., 1996. Water waves generated by underwater explosion. Adv. fatalities due to tsunamis (Mastin and Witter, 2000). Hazard of volcanic Ser. Ocean Eng. World Sci, New Jersey. tsunamis in lakes is particularly concerning and somewhat neglected Mastin, L.G., Witter, J.B., 2000. The of eruptions through lakes and seawater. J. Volcanol. Geotherm. Res. 97, 195–214. (Freundt et al., 2007; Paris, 2015) if we consider that the coasts of many May, S.M., Falvard, S., Norpoth, M., Pint, A., Brill, D., Engel, M., Scheffers, A., Dierick, M., volcanic lakes in the world are densely populated (e.g. Lake Nicaragua Paris, R., Squire, P., Brückner, H., 2015. A mid-Holocene candidate tsunami deposit from the NW Cape (Western Australia). Sediment. Geol. 332, 40–50. and Lake Managua in Nicaragua, in the Democratic Republic Moore, A., Goff, J., McAdoo, B.G., Fritz, H.M., Gusman, A., Kalligeris, N., Kalsum, K., of Congo, Lake Taal in the Philippines). Susanto, A., Suteja, D., Synolakis, C.E., 2011. Sedimentary deposits from the 17 July 2006 Western Java tsunami, Indonesia: use of grain size analyses to assess tsunami flow depth, speed, and traction carpet characteristics. Pure Appl. Geophys. 167 (11), Acknowledgments 1951–1961. Morrissey, M., Gisler, G., Weaver, R., Gittings, M., 2010. Numerical model of crater lake – This study was funded by the Conseil Régional d'Auvergne, (project eruptions. Bull. Volcanol. 72, 1169 1178. Morton, R.A., Gelfenbaum, G., Jaffe, B.E., 2007. Physical criteria for distinguishing sandy TSUNAMIX – grant 79285), and by the French National Research tsunami and storm deposits using modern examples. Sediment. Geol. 200 (3–4), Agency, (project VITESSS – grant ANR08-JCJC-0042) (Volcano-Induced 184–207. Morton, R.A., Goff, J.R., Nichol, S.L., 2008. Hydrodynamic implications of textural trends Tsunamis: Sedimentary Signature and numerical Simulation). The in sand deposits of the 2004 tsunami in Sri Lanka. Sediment. Geol. 207 (1–4), 56–64. authors are particularly grateful to Elisabeth Miot-Noirault (INSERM, Muravyev, Y.D., Fedotov, S.A., Budnikov, V.A., Ozerov, A.Y., Magus'kin, M.A., Dvigalo, Clermont-Ferrand), Paul Tafforeau and Elodie Boller (ID 19 ESRF, V.N., Andreyev, V.I., Ivanov, V.V., Kartasheva, L.A., Markov, I.A., 1998. Volcanic activity in the Karymsky center in 1996: summit eruption at Karymsky and Grenoble), Isabelle Billy (EPOC CNRS, Bordeaux), to all the participants phreatomagmatic eruption in the Akademii Nauk caldera. Volcanology and of the survey to Karymskoye Lake, and to Witold Szczuciński and an Seismology 19, 567–604. anonymous reviewer. This is ClerVolc Laboratory of Excellence con- Narayana, A.C., Tatavarti, R., Shinu, N., Subeer, A., 2007. Tsunami of December 26, 2004 on the southwest coast of India: post-tsunami geomorphic and sediment tribution number 251. characteristics. Mar. Geol. 242, 155–168. Paris, R., 2015. Source mechanisms of volcanic tsunamis. Phil. Trans. R. Soc. A 373, 20140380. References Paris, R., Wassmer, P., Lavigne, F., Belousov, A., Belousova, M., Iskandarsyah, Y., Benbakkar, M., Ontowirjo, B., Mazzoni, N., 2014. Coupling eruption and tsunami Auker, M.R., Sparks, R.S.J., Siebert, L., Crosweller, H.S., Ewert, J., 2013. A statistical records: the Krakatau 1883 case-study, Indonesia. Bull. Volcanol. 76, 814. ff ff analysis of the global historical volcanic fatalities record. J. Appl. Volcanol. Pilarczyk, J.E., Horton, B.P., Witter, R.C., Vane, C.H., Chagué-Go , C., Go , J., 2012. 2013 (2:2). Sedimentary and foraminiferal evidence of the 2011 Tohoku-oki tsunami on the – Babu, N., Babu, D., Das, P., 2007. Impact of tsunami on texture and mineralogy of a major Sendai coastal plain, Japan. Sediment. Geol. 282, 78 89. placer deposit in southwest coast of India. Environ. Geol. 52, 71–80. Srinivasalu, S., Thangadurai, N., Jonathan, M.P., Armstrong-Altrin, J.S., Ayyamperumal, Bahlburg, H., Weiss, R., 2006. Sedimentology of the December 26, 2004 Sumatra tsunami T., Ram Mohan, V., 2008. Evaluation of trace-metal enrichments from the 26 deposits in eastern India (Tamil Nadu) and Kenya. Int. J. Earth Sci. 96 (6), December 2004 tsunami sediments along the Southeast coast of India. Environ. Geol. – 1195–1209. 53, 1711 1721. Belousov, A., Belousova, M., 2001. Eruptive process, effects and deposits of the 1996 and Switzer, A., Pucillo, K., Haredy, R., Jones, B., Bryant, E., 2005. Sea level, storm, or ancient basaltic phreatomagmatic eruptions in Karymskoye lake, Kamchatka, Russia. tsunami: enigmatic sand sheet deposits in a sheltered coastal embayment from – In: White, J., Riggs, N. (Eds.), Volcaniclastic Sedimentation in Lacustrine Settings. southeastern New South Wales, Australia. J. Coast. Res. 21, 655 663. 30. Blackwell, International Association of Sedimentologists Special Publications, pp. Switzer, A., Srinivasalu, S., Thangadurai, N., Mohan, W.R., 2012. Bedding structures in 235–260. Indian tsunami deposits that provide clues to the dynamics of tsunami inundation. – Belousov, A., Belousova, M., Muravyev, Y., 1997. Holocene eruptions in Akademiya Nauk Geol. Soc. Lond., Spec. Publ. 361, 61 77. ń caldera and the age of Karymsky stratovolcano, Kamchatka. Transactions of the Szczuci ski, W., Chaimanee, N., Niedzielski, P., Rachlewicz, G., Saisuttichai, D., Russian Academy of Sciences, Earth Sciences Section 335, 653–657. Tepsuwan, T., Lorenc, S., Siepak, J., 2006. Environmental and geological impacts of — Belousov, A., Voight, B., Belousova, M., Muravyev, Y.D., 2000. Tsunamis generated by the 26 December 2004 tsunami in coastal zone of Thailand overview of short and ff – subaquatic volcanic explosions: unique data from 1996 eruption in Karymskoye lake, long-term e ects. Pol. J. Environ. Stud. 15 (5), 793 810. ń ń ff Kamchatka, Russia. Pure Appl. Geophys. 157, 1135–1143. Szczuci ski, W., Kokoci ski, M., Rzeszewski, M., Chagué-Go , C., Cachao, M., Goto, K., Cheel, R.J., 1984. Heavy mineral shadows, a new sedimentary structure formed under Sugawara, D., 2012. Sediment sources and sedimentation processes of 2011 Tohoku- — upper flow regime conditions: its directional and hydraulic significance. J. Sediment. oki tsunami deposits on the Sendai Plain, Japan insights from diatoms, nannoliths – Petrol. 54, 1173–1180. and grain size distribution. Sediment. Geol. 282, 40 56. Cheel, R.J., 1990. Horizontal lamination and the sequence of bed phases and stratification Torsvik, T., Paris, R., Didenkulova, I., Pelinovsky, E., Belousov, A., Belousova, M., 2010. under upper-flow-regime conditions. Sedimentology 37, 517–529. Numerical simulation of explosive tsunami wave generation and propagation in – Costa, P.J.M., Andrade, C., Cascalho, J., Dawson, A.G., Freitas, M.C., Paris, R., Dawson, S., Karymskoye Lake, Russia. Nat. Hazards Earth Syst. Sci. 10 (11), 2359 2369. 2015. Onshore tsunami sediment transport mechanisms inferred from heavy mineral Ulvrová, M., Paris, R., Kelfoun, K., Nomikou, P., 2014. Numerical simulations of tsunami assemblages. The Holocene 25 (5), 795–809. generated by underwater volcanic explosions at Karymskoye Lake (Kamchatka, Craig, B.G., 1974. Experimental Observations of Underwater Detonations Near the Water Russia) and Kolumbo volcano (Aegean Sea, Greece). Nat. Hazards Earth Syst. Sci. 14, – Surface. Los Alamos Scientific Laboratory of the University of California (report LA- 401 412. 5548-MS, UC-34). Wohletz, K.H., 1986. Explosive magma-water interactions: thermodynamics, explosion fi – Cuven, S., Paris, R., Falvard, S., Miot-Noirault, E., Benbakkar, M., Schneider, J.-L., Billy, mechanisms, and eld studies. Bull. Volcanol. 48, 245 2654.

170